Coppermine vs Celermine at Seti

[blade47]

With v2.4 the Coppermine destroyed the Celermine at seti. Now that cache and bus speed are less important how do the C2's compare to the P3's?

I've seen many people posting good times with the C2 and new client. What I really would like to know is on a clock for clock basis how do the two stack up using v3.0?

 

[poopaskoopa]

I've been running C2 566@850 with 256mb ram (pc100 generic) at 2-2-2. I think my best time is right around 5hrs(3.0 CLI). I'm pretty sure, but not absolutely certain, since I'm never around when it finishes a WU. I think the time with 3.0 CLI varies between 5 ~ 7 hrs on C2 850 machine too.

 

[vss1980]

Technically, the standard PIII-Coppermine should be faster than an equivalently clocked Celermine as it has more cache, and that the cache is supposedly better optimised (PIII has a higher X-way associative cache).

 

[Lvis]

My 566@566 would average a about ten and a half hours with the old client. Now I seem to be doing them in about nine and a half with 3.0. I also switched to win me at the same time, I don't know if that's a factor or not.

 

[BurntKooshie]

VSS - that's exactly the idea. The Coppermine has a higher hit-rate for two reasons: one, it has a larger cache (more likely to hold the important information), and: two, because, given the same latencies (such as exists between the Coppermine and the Culeron), a lower associativity decreases hit rate given the same cache size (its not incredibly drastic, but it is there).

Basically, the Culeron is worse off because of two things: A) the hit rate (due to cache size / associativity), and B) the memory bandwidth. However, what most people don't understand is that the Culeron 850 (overclocked of course) has the exact same memory bandwidth from the main memory as a theoretical PIII 850. So yes, the 850 Culeron has less main memory bandwidth than does a PIII 800B, but then again, so does a PIII 800 (not 133mhz bus)

 

[Hellburner]

ARS v3.0 Benchmarks

Based upon the above and my own numbers the Celeron 2 is 17% slower than a Pentium 3e at the same clock speed and roughly same FSB. The effects of FSB and memory timing have not been looked into much, I'm sure they have some sort of effect but they are far subtler than they use to be. Maybe I could check it out tonight.

 

[Hellburner]

I checked out the memory timimg, it's definitely still a factor. On a P3 650@866 there is a 14% slow down when you drop from CLT 3 Cas 2-2-2 to CLT 4 Cas 3-3-3.

I'm still working on the FSB factor.

 

[blade47]

Thanks Hellburner.

 

[Assimilator1]

Thanks HB ,I guess I could experiment with FSB on my PII system ,eg 280MHz ,2.5 x112 vs 293MHz 3.5 x 83MHz.

 

[blade47]

Hey Assim1, on the slower cpus bus speed isn't the bottleneck so it doesn't really help that much. An example is my P2-233.
280mhz=112*2.5
is slower than
290mhz=83*3.5
with both v2.4 and v3.0.

Luckily my P2-233 will do 309mhz easily. Now if my mobo would just let me bump the voltage a little I could be running it at 351mhz.

 

[BurntKooshie]

blade - that's due to the fact that the multipliers are still relatively low. Its more the ratio of the core speed to the bus speed that is important, as I bet if you scaled the times of a culeron 566 with 66mhz bus to a culeron 850 on a 100mhz bus, I bet the ratio of times would be rather close to the ratio of the core:bus.

 

[Assimilator1]

Blade_47

351MHz ???:Q you'll be lucky! ,my PII 233 is unstable at 300 MHz!
I might try 293Mhz ,though ,then again I could get a Cel 300 @ 450 for it

 

[blade47]

Assim1, mine will do 351mhz I've run it before in my BX6rev2.0 mobo. All I've got to do is bump the voltage to 2.9v (2.8v is the default). Unfortunately my Gigabyte 6BXC mobo I'm running the P2-233 in doesn't have voltage adjustment and the only info I could find on pin taping was for 2.0-2.2v.

BTW have you decased your P2-233? Mine wouldn't even run stable at 290mhz. It would load into windows and then I'd get random errors. I popped the case off (which was very easy), cleaned off the ton of goop intel calls thermal grease and reapplied a thin layer of fresh stuff & put the stock heatsink and fan back on and voila 309mhz and it runs cooler than it did before at 262mhz.

Of course then I had to see how high it would go. So I dropped it in my BX6 and bumped the voltage and it chugged along at 351mhz (3*117) with ease.